AMD Lisa Su on AI: Why Fast Experimentation Beats Perfect Planning

May 22 2025 10:58
Under Lisa Su's leadership, AMD has grown from a $4 billion company to over $23 billion, transforming into one of the world's fastest-growing semiconductor businesses. But perhaps more telling than these numbers is how Su approaches the technology that's reshaping every industry: AI. In a recent conversation with HBR editor Adi Ignatius, Su shared insights that cut through the AI hype to offer practical wisdom for business leaders navigating this transformative moment.

Her message is refreshingly pragmatic. While others debate AI's existential threats or promise utopian futures, Su focuses on what works now and what leaders can do today to harness AI's potential responsibly.


Su describes herself as a "techno optimist" but with an important caveat: she's pragmatic about it. "The technology is not perfect," she acknowledges. "We're still in the very early stages of the deployment of AI, and we do know that the AIs are not always right."

This perspective shapes everything AMD does with AI. Rather than waiting for perfect solutions or fearing worst-case scenarios, Su advocates for active learning and responsible experimentation. "What we have to do as leaders of companies is to really learn how to harness the power of AI and also bring our employees along with that so that we're actually making our employees more productive."

The key insight here isn't about the technology itself but about leadership mindset. Su sees AI adoption as fundamentally about making people more capable, not replacing them. This human-centered approach to AI deployment offers a middle path between uncritical enthusiasm and paralyzing caution.


When asked about balancing speed versus caution in bringing AI products to market, Su's answer is unequivocal: "I really believe in fast experimentation and implementation. I don't believe the answer is let's slow down."

But speed doesn't mean recklessness. AMD has established a Responsible AI Council to guide their efforts, focusing on protecting intellectual property for both the company and its partners. The real innovation lies in Su's approach to finding high-value use cases quickly.

"I think the power of AI is finding those use cases that give you very, very significant return on investment," Su explains. At AMD, this philosophy has transformed their design workflows. "We've seen what used to take weeks and months really come down to days."

This dramatic time compression isn't just about efficiency. It represents a fundamental shift in how the company operates and competes. When your design cycles compress from months to days, you can iterate faster, respond to market changes more quickly, and ultimately deliver better products.


Su's advice for leaders struggling to move beyond AI experimentation to actual implementation is remarkably concrete. She recommends starting with "copilot" applications where AI assists rather than replaces human expertise. At AMD, these copilots appear across multiple functions:

• Engineering: AI helps write code, examine test cases, and improve quality assurance processes
• Marketing: AI assists with content creation and communications, with human experts providing final touches
• Business intelligence: AI analyzes sales cycle patterns and provides predictive insights for strategic planning

The pattern here is instructive. AMD doesn't use AI to fully automate complex tasks but to accelerate the path to answers. As Su puts it, these copilots "allow you to get close to the answer, and then of course the final touches are being done by your expert employees."

This approach reduces risk while maximizing value. Human expertise remains central, but AI dramatically increases productivity and speed.


Su's emphasis on responsible AI isn't just about ethics, it's about sustainable business practice. AMD's Responsible AI Council focuses on practical concerns that every business faces: protecting intellectual property, ensuring customer data security, and maintaining quality standards.

But responsibility also means realistic expectations. "There are places where of course you have to be a little bit more careful," Su notes, "places that you would rely more on the AI itself to come up with the answer. And there you have to do a lot of testing to make sure that you get the right answers."

1. Start with lots of pilots and experimentation
2. Focus on areas with high value and low barriers to entry
3. Invest more heavily in training AI for complex, business-specific use cases
4. Always maintain human oversight for critical decisions

This approach allows companies to gain AI experience and confidence while minimizing risks and building internal capabilities.


Su's decade as AMD's CEO offers lessons that extend far beyond AI. When she took over, AMD faced the classic challenge of limited resources and unlimited opportunities. Her solution was ruthless focus on what AMD could do better than anyone else: high-performance computing.

This focus proved prescient. By doubling down on high-performance computing before it became central to AI, AMD positioned itself perfectly for the current boom. "Between high-performance computing and AI, we are in perhaps one of the most exciting areas, if not the most exciting area in semiconductors."

The lesson for other leaders is clear: success often comes from saying no to good opportunities so you can say yes to great ones. In a world where AI creates new possibilities daily, this kind of strategic focus becomes even more critical.


One of the most practical insights from Su's interview addresses a concern many business leaders share: will AI be too expensive for smaller companies to access? Su's answer reveals how technology leaders think about market development.

"The great thing about technology, especially when you think about usage curves, is we're very cognizant of the fact that for technology to be most broadly adopted, you do actually need to get the cost to a very, very reasonable point."

AMD is actively working to reduce AI costs "by factors over the next couple of years." This isn't just about hardware pricing, but about making AI inference operations, like asking questions to chatbots, dramatically cheaper.

Su's perspective here is both strategic and democratic. Widespread AI adoption benefits companies like AMD because it creates larger markets. But it also reflects a genuine belief that AI's benefits shouldn't be limited to the largest companies with the biggest budgets.


Perhaps the most valuable insight from Su's interview comes when she's asked about the most important lesson from her decade as CEO. Her answer cuts to the heart of effective leadership in fast-moving industries.

This balance between long-term vision and short-term execution is particularly relevant for AI adoption. The technology's potential is enormous, but realizing that potential requires patient, sustained effort combined with rapid experimentation and clear progress markers.

Under Su's leadership, AMD grew from $4 billion to over $23 billion in revenue. This wasn't luck or market timing alone, though both played roles. It was the result of setting ambitious goals, making focused bets, and executing consistently over time.

For business leaders thinking about AI strategy, Su's approach offers a practical framework. Set ambitious goals for what AI could do for your business in three to five years. Then break that vision down into quarterly and annual milestones. Start experimenting now, but always with an eye toward longer-term strategic advantage.